Reproducibility in data sciences: why, what, and how
An invited talk at the University of Zurich on the background, motivations and methodology of reproducible research.
One of the key principles of proper scientific procedure is the act of repeating an experiment or analysis and being able to reach similar conclusions. Published research based on computational analysis, for example in bioinformatics or computational biology, has often suffered from incomplete method descriptions (such as the list of software versions used), unavailable raw data, and incomplete, undocumented or unavailable code. This essentially prevents any attempt to reproduce the results of such studies. The term reproducible research describes the idea that a scientific publication based on computational analysis should be distributed along with all the raw data and metadata used in the study, all the code and computational notebooks needed to produce results from that raw data, and the computational environment or a complete description of it.
Reproducible research not only leads to proper scientific conduct, it also gives other researchers the access needed to build on previous work. Most importantly, whoever sets up a reproducible research project quickly realises the immediate personal benefit: an organised and structured way of working. The person who most often has to reproduce an analysis is one’s own future self.
The talk covers the key motivating factors and issues in building reproducible code, and ends by showcasing tools and a strategy for building a reproducible research ecosystem.
- Slides (PDF)